Otherwise, transactions from client and from server could be out of order in the undo/redo stack, meaning the latest action might not be undo-able without undoing the second-to-last action first, which is, of course, undesirable.
If all writes are done by server code (as planned), this issue goes away.
This is a temporary fix until the babrahams:transactions2 package comes out, which will make all writes via a method, meaning we'll only ever be dealing with server time.
Otherwise, transactions from client and from server could be out of order in the undo/redo stack, meaning the latest action might not be undo-able without undoing the second-to-last action first, which is, of course, undesirable.
If all writes are done by server code (as planned), this issue goes away.